The Agent Skills Directory

[SAFE]: The skill is entirely instructional, providing a guide for creating other skills following progressive disclosure and modularity principles.
[INDIRECT_PROMPT_INJECTION]: The guide mentions workflows for processing external data formats such as PDF, Excel, and Word. While this identifies a potential injection surface for skills built using these patterns, the instructions are educational and recommend validation and verification steps to improve reliability.
Ingestion points: references/best-practices.md describes methods for extracting text from external documents.
Capability inventory: The skill discusses executing Python and Node.js scripts for data analysis.
Boundary markers: The guide encourages the use of clear workflows and feedback loops.
Sanitization: The skill emphasizes the use of validation scripts and error handling.
[DYNAMIC_EXECUTION]: The skill provides examples of script generation and execution (e.g., using Python for PDF extraction or Node.js for Word processing). These patterns are presented as best practices for achieving deterministic results in agent workflows and do not involve untrusted code execution within this skill.

create-skill